R Notebook

library(tidyverse)
library(caret)
library(lattice)
library(DataExplorer)

# Get barstool data off github
job_post <- read_csv("https://raw.githubusercontent.com/luizmalpele/datasets/master/fake_job_postings.csv")

Warning: closing unused connection 3 (https://raw.githubusercontent.com/luizmalpele/StatsLearningProject/master/data/fake_job_postings.csv)
Parsed with column specification:
cols(
  job_id = col_double(),
  title = col_character(),
  location = col_character(),
  department = col_character(),
  salary_range = col_character(),
  company_profile = col_character(),
  description = col_character(),
  requirements = col_character(),
  benefits = col_character(),
  telecommuting = col_double(),
  has_company_logo = col_double(),
  has_questions = col_double(),
  employment_type = col_character(),
  required_experience = col_character(),
  required_education = col_character(),
  industry = col_character(),
  `function` = col_character(),
  fraudulent = col_double()
)

job_post

Plot Missing Values

plot_missing(job_post)

Missing Information Comparison

job_post %>%
  select(fraudulent, department, required_education, benefits, required_experience,salary_range, location, requirements, company_profile, employment_type, industry) %>% 
  group_by(fraudulent) %>% 
  summarize(na_ratio_salary = sum(is.na(salary_range))/length(salary_range),
            na_ratio_department = sum(is.na(department))/length(department),
            na_ratio_required_education = sum(is.na(required_education))/length(required_education),
            na_ratio_benefits = sum(is.na(benefits))/length(benefits),
            na_ratio_requirements = sum(is.na(requirements))/length(requirements),
            na_ratio_company_profile = sum(is.na(company_profile))/length(company_profile),
            na_ratio_location = sum(is.na(location))/length(location),
            na_ratio_employment_type = sum(is.na(employment_type))/length(employment_type),
            na_ratio_industry = sum(is.na(industry))/length(industry)
            )

The variables that presented higher missing information ratio are: company_profile and employment_type.

job_post %>% group_by(fraudulent) %>% 
  summarize(ratio_has_questions = sum(has_questions)/length(has_questions),
            ratio_has_company_logo = sum(has_company_logo)/length(has_company_logo),
            ratio_telecommuting = sum(telecommuting)/length(telecommuting)) 

The variables that presented higher missing information ratio are: ratio_has_questions and ratio_has_company_logo. The next step is to investigate the titles and decriptions using data mining and text mining tecniques.